After building the graph, we have to execute it, right? For computing a graph, we should put it in a session. A Session class in TensorFlow places all ops or nodes onto computational devices such as CPU or GPU.

Here is how we create a Session object in TensorFlow:

    sess = tf.Session() 

For running the ops in a graph, the Session class provides methods to run the entire graph:

    print(sess.run(hello)) 

It will execute the op called hello and print "Hello, TensorFlow" in Terminal.